Intel will have around $120M external foundry revenue in CY25 or 1000x less than TSMC. While 18A didn't get foundry traction in the initial wave, Intel still pins hopes on overall foundry (internal plus external) and expects it to break-even exiting '27, even with low contribution from external customers. As Intel moves its internal wafers to EUV from low-base (5% EUV penetration in CY24), they expect wafer ASP to grow 3× faster than costs, narrowing losses.

SanDisk's capital intensity will be mid-to-high teens in the near term vs mid-teens long term guidance. NAND WFE capex peaked in 2021 at $20B+ but current run-rate is $10B+. SanDisk suggests unconstrained NAND demand at mid-20% bit growth for CY26 but supply will be limited at 17%. With DC overtaking smartphone in end demand and fabs running at 100% utilization, NAND investment likely pick up. But how much is still a question after last cycle's missteps. LTAs won't work well in memory.

For the first, QCOM disclosed their QCT COGS explicitly. It doesn't take a genius to calculate though, given QTL IP revenue has 95% GM.

Qualcomm's revenue from Apple and Samsung grew below corporate average in FY25 while China outgrew. Apple revenue recovered in iPhone 16 cycle (X75 modem) after declining in iPhone 15 cycle. Qualcomm still maintains 20% share in FY26 launches (no change).

- Days of inventory down to 199 days from 214 days in QJun - Distributor inventory was at 27 days, down 2 days q/q - Lead times are increasing for some products (substrates, subcontracting capacity, and some foundry constraints) - 3nm PCIe Gen 6 switch product announcement; claims industry first 3nm; volume ramp by the end of CY26; $2B TAM growing at 10% CAGR through 2030 - targeting hyperscalers, enterprise OEMs, and ODMs, covering all segments of the AI and enterprise data center markets

Microchip QSep25 - Revenue up 6% q/q and xx% y/y to $1.14B at 55.9% GM; Asia and Americas up and Europe flat; QDec, a seasonally weak quarter, will be down 1% q/q - expects strong quarters for Mar, Jun and Sep 2026; strong backlog trend for QMar vs QDec - Underutilization and write-offs affected 10.8 percentage points of non-GAAP GM - Seeing recovery in key end markets - DC had the strongest performance last quarter - The book-to-bill ratio for the last quarter was 1.06

- Banking on bit growth from BiCS8 - Long term supply agreements with customers (unheard of); but admits as of today very few volume and price commitments extend beyond a single quarter - Capex plans remain unchanged for FY26 - HBF controller in ’27 - Faster growth of capacity in PC and smartphone vs unit growth

- not currently considering adding new wafer capacity; 100% utilization - Bit shipments up mid-teens while ASPs up MSD% - QDec to see MSD% bit growth and DD% price increases - Claims share gains in DC - QLC is projected to grow from 20% to 40% of SNDK’s enterprise business by the end of FY26 - BiCS8 15% of bit production in QSep and will be the majority by the end of FY26 - DC business up 26% q/q; DC up 2ndstraight quarter after bottoming out in QMar

SanDisk QSep25 - Revenue up 21% q/q and 23% y/y to $2.3B; $2.6B guidance for QDec - Claims DC will overtake smartphone as the largest end market for NAND in ‘26 - increased its forecast for DC exabyte growth in calendar 2026 from a mid-20% level to a mid-40% level - Demand outpacing supply; NAND market likely constrained through 2026 and beyond, per SNDK

ClusterMAX 2.0: The Industry Standard GPU Cloud Rating System 95% Coverage By Volume, 84 Providers Rated, 209 Providers Tracked, 140+ Customers Surveyed, 40,000 Words For Your Enjoyment newsletter.semianalysis.com/…

Non-Apple semiconductor revenue growing faster than Apple revenue. Auto and IoT at this growth rate can offset a good chunk of lost Apple revenue readily.

- Licensing revenue $1.4B; will grow slightly to $1.5B (mid-point guidance) in QDec - 18% y/y growth in non-Apple semi revenue - non-Apple QCT revenues grew at 5 yr CAGR of 15% through FY25 - 150 PC designs to be commercialised through ’26 (100+ in ’26 alone)

- Says multiple billions of DC revenue 2 yrs from now; revenue starts from FY27 (starts from Oct 26) - 2 areas of focus within DC: CPU and inference - Sticks to its non-HBM DC inference approach; says it is differentiated in compute density and power consumption - Alphawave closure in 1Q26 - DC opex incremental to spending profile; shifting opex dollars from mature businesses such as handsets

- Auto crosses $1B mark first time (up 7% q/q and 17% y/y) and will stay at this level for the next quarter too; on track to $8B auto revenue by FY29; expects auto revenue to grow through FY26 - ADAS stack co-designed with BMW now available in 60 countries and will extend to 100 in ‘26 - Reiterates FY29 non-handset target ($22B); DC will be incremental

- Targets 75% baseline share for S26; been saying this for a while - No change in 20% Apple share assumption in FY26 launches - No pull forward from China - IoT revenue up 8% q/q and 7% y/y to $1/8B; driven by AI smart glasses - IoT will decline in QDec driven by consumer seasonality - IoT will grow in FY26 vs FY25

- Handset revenue to grow low-teens % q/q in QDec; most of the incremental from Android though some benefit from Apple as well - Apple (21%), Samsung (20%) and Xiaomi (13%) top-3 customers in FY25 (my interpretation) - Premium tier expansion and content growth driving handset revenue - OEM competition and consumer desire to upgrade factors as well in premium tier growth; sees this as a sustainable trend

Qualcomm QSep25 - Revenue up 9% q/q and 10% y/y to $11.3B (above the high-end guidance); $11.8-$12.6B guidance for QDec - QCT revenue up 9% q/q and 13% y/y to $9.8B at 29% EBT; all end markets up q/q; $10.3-$10.9B guidance for QDec - Inflection point in smart glasses; for the 2nd straight quarter calls out it as a growth driver; design-wins in Meta; Ahead of targets in auto and XR - Handset revenue of $7B up 10% q/q and 14% y/y; driven by premium tier Android and new Snapdragon 8 Elite 5

- +41% q/q jump in related party revenue driven by SoftBank licenses; $178M; will be the run-rate going forward; license plus design services; the latter is low margin; Stargate related revenue - The company will talk about own chips once products taped out, samples out from the fab and it secures non-cancellable orders - Expects Cloud royalties (which includes server CPU) contribution to go up to 15-20% in FY26 vs 10% in FY25 - China demand strong, driven by licensing; one of largest licensing deals

- ACV up 28% y/y; well above licensing long-term CAGR of mid-to-high single-digit - Increased R&D to fund compute SoCs as Arm ventures into chips - Smartphone Lumex CSS had 450 people and 100s of million of investment; early revenue started - Ethernet IP acquisition from DreamBig Semiconductor

- DC Neoverse core royalties up >2x y/y - Licensing up 56% y/y to $515M - 3 new CSS licenses (1 each in smartphone, tablet and DC); total 19 CSS licenses across 11 companies - Samsung Exynos on CSS - Claims top-4 Android companies shipping CSS devices; CSS offers 2x royalty vs v9